Let me tell you what makes my blood boil.
An account lands on my desk. Gorgeous ROAS graph, 7x, up and to the right, previous agency taking bows on the way out. And the owner sits across from me and asks the question the graph can’t answer:
“So why am I not making any money?”
Because ROAS treats every rand of revenue as equal.
And it isn’t. Not even close.
The math your report doesn’t show
Two products, both selling at R1,000.
Product A carries 10% margin. R100 in your pocket per sale.
Product B carries 60%. R600 per sale.
Now run both at a 5x ROAS:
A: R5,000 revenue → R500 gross profit
B: R5,000 revenue → R3,000 gross profit
Same ROAS. Six times the money.
Hand that account to Google’s bidding algorithm with a ROAS target and no margin data, and it will hunt whatever converts cheapest against revenue. Which is usually your discounted, low-margin, gets-returned-half-the-time stock. The dashboard says winning.
Your bank account says otherwise.
Three teams, one week, same answer
Here’s what got me writing this. In a single week, three completely unconnected sources told me the same thing in different words.
Common Thread Collective, the US agency buying ads for 170-plus e-commerce brands, breaking down how they structure Google accounts. ProfitMetrics, whose entire product is piping profit data into ad platforms. And a paid-search newsletter from Echelonn walking through margin-structured Shopping campaigns, case-study numbers attached.
Our knowledge system has a rule: nothing gets into the playbook until independent sources converge on it. Most ideas die at one mention. This one hit three in five days.
When that happens, I sit up.
The shared answer: stop feeding the algorithm revenue. Feed it margin. POAS, profit on ad spend, if you want the acronym.
Two moves follow.
One. Structure Shopping and Performance Max by margin band, not product category. High-margin stock gets its own campaign and its own target. The algorithm can’t subsidise the rubbish with the good stuff anymore, because they’re not in the same bucket.
Two, and this is the one that sounds wrong until it clicks: lower the target. Once profit is what you’re optimising, a softer target gives the algorithm more room in the auctions on stock you actually make money on. Volume goes up, and every extra sale is profitable by construction. That high ROAS target you were so proud of? It was the thing choking growth. Echelonn’s case numbers show exactly this shape.
So why isn’t your agency doing this?
Honestly? Because ROAS is easy and margin data is work. It means asking your finance person for cost of goods, and some unglamorous plumbing in a product feed. A 7x screenshot fits in a report. “We made you more actual profit at a lower ROAS” takes ten minutes to explain.
Once. It takes ten minutes once.
The metric you optimise is the business you build. Optimise revenue and you will scale revenue, including the revenue that loses you money. I’ve watched it happen to good businesses run by smart people, and it never stops making me angry.
If your agency has never asked for your margins, they are not optimising your profit. They can’t be. They don’t have the number.
Ask them for your POAS this week.
Watch the face.